Why Violence and AI Are Both Governance Problems Before They Are Technology Problems

May 07, 2026

9 min read

5 views

84%

The surprising common denominator: breakdowns in governance of human attention

What do political violence, online radicalization, police escalation, and artificial intelligence have in common? At first glance, almost nothing. One belongs to the realm of streets, rallies, and public fear. The other belongs to algorithms, regulations, and the architecture of digital systems. But the deeper connection is not about subject matter. It is about how modern societies govern attention, belonging, and escalation.

That is the uncomfortable insight hiding beneath both debates. Political violence does not begin with violence. It begins with a distorted social environment in which people misread what others believe, feel morally outraged, search for status or meaning, and find those impulses amplified by systems that reward provocation. AI governance enters the picture because the same pattern appears in digital infrastructure: tools that shape what people see, how fast they react, and how likely they are to treat others as enemies. The issue is not simply bad actors or bad software. It is the failure to build institutions that slow escalation before it becomes action.

The core challenge of our era is not merely preventing harm. It is designing environments that do not constantly manufacture the conditions for harm.

That is why political violence and AI belong in the same conversation. Both expose a society that is technologically powerful but civically underdeveloped.

Escalation happens when perception outruns reality

One of the most counterintuitive findings in the political violence research is that most people reject political violence, yet they overestimate how much their opponents support it. This matters because violence is rarely justified by raw belief alone. It is justified by a story: the story that the other side is not just wrong, but dangerous, irredeemable, and already committed to harm.

That story spreads more easily in environments built for outrage. Algorithms prioritize negative, hostile, and identity-threatening content because those posts keep attention. Human brains, especially under stress, are quick to mistake repetition for consensus. If every feed item says the other camp is vicious, a person may conclude that violence is common, even when polling says it is almost universally rejected. This is not just misinformation. It is perceptual distortion at scale.

The same logic helps explain why AI governance matters. AI systems increasingly mediate what we think is important, urgent, and normal. Recommendation engines, ranking systems, and generative tools can compress nuance, intensify tribal language, and make extreme interpretations feel ambient. Even when the model is not explicitly political, its architecture can create the emotional weather in which political radicalization becomes easier.

This is the first bridge between the two themes: when systems distort perception, they reduce the gap between irritation and action. The danger is not only what people believe. It is how quickly they believe that their beliefs are shared, justified, and imminently threatened.

The real fuel is often moral identity, not ideology

A shallow reading of political violence treats extremists as ideologues with rigid platforms. But the highlights point to a more useful model. Many people who cross the threshold into violence are not simply committed to a coherent doctrine. They are action oriented, driven by moral outrage, grievance, status seeking, alienation, and a desire to matter. Some are “salad bar extremists,” assembling beliefs from disparate sources. Others are pulled by trait victimhood, a persistent tendency to interpret the world through the lens of being wronged.

This is crucial because it changes the prevention problem. If violence were mainly a matter of ideology, the answer would be better arguments. But if it is often a matter of wounded identity, then the deeper needs are belonging, recognition, efficacy, and meaning. That is why deradicalization efforts that address victimhood, provide alternative groups, or redirect sensation seeking can work better than purely informational interventions.

Here, AI reappears not as a separate topic, but as an accelerant. AI systems can intensify identity loops by continuously tailoring content to the user’s resentments, curiosities, and insecurities. A person searching for dignity can be led toward a digital ecosystem that rewards outrage, confirms persecution narratives, and supplies a ready-made enemy. In that sense, AI is not only a tool of persuasion. It is a factory for narrative reinforcement.

A useful mental model is this:

Violence often emerges when moral injury, social isolation, and a sense of powerless agency meet a system that repeatedly says, “Yes, your enemies are real, and yes, you are justified.”

That is not an ideology. That is a feedback loop.

The most important prevention tool is not force, but design

A society’s reflex in the face of violence is often to harden. More surveillance, more force, more compliance, more control. But the political violence research warns that heavy-handed policing can trigger the very escalation it seeks to contain. When authorities rely on militarized or violent responses, they may generate a sense of humiliation, deepen grievances, and transform protests into proof of oppression. The result is what some call tyrannical peace: order imposed in ways that make future disorder more likely.

This is where the connection to AI governance becomes especially revealing. Effective AI regulation is not only about banning egregious uses after they appear. It is about shaping the incentives and defaults of systems before harm becomes entrenched. In both domains, the best intervention is often not a dramatic crackdown but a design choice:

In civic life, procedural justice and fair treatment reduce the likelihood that people interpret authority as domination.
In digital life, safer recommendation systems and transparent safeguards reduce the likelihood that engagement incentives become escalation incentives.

Think of it like traffic engineering. A city can respond to accidents by punishing drivers more harshly, or it can redesign intersections so crashes are less likely in the first place. Most modern harm is intersection harm. It happens where human weakness meets system design.

The same is true for violent radicalization. People often leak their intent long before acting: to friends, family, coworkers, peers online. Threat assessment tools, anonymous hotlines, and student awareness campaigns work not because they create omniscience, but because they turn communities into sensors. Prevention becomes possible when ordinary people know what to look for and feel safe reporting it.

That insight should also shape AI governance. The point is not to imagine a single master regulator who catches every risk. The point is to create distributed systems of early warning, accountability, and humane intervention. Governance is not just law. It is architecture.

Belonging is the hidden variable in both democracy and AI

If there is one concept that threads through all of this, it is belonging. People drawn toward violence often feel excluded, disrespected, or invisible. Some seek status, some seek revenge, some seek protection for an imagined in-group. Even police officers, under certain institutional norms, can dehumanize protesters when those protesters seem to threaten authority. The common factor is that once people stop seeing others as full participants in a shared world, escalation becomes easier.

This is why the community-based prevention approach matters so much. Programs that improve emotional wellness, cognitive flexibility, fairness, justice, reduced problematic internet use, and belonging without othering are not soft extras. They are the infrastructure of democratic stability. They widen the space between frustration and harm.

Now connect that to AI. One of the most neglected social effects of AI is that it can either strengthen or weaken belonging. If it becomes a machine for isolation, hyper-personalized grievance, and constant comparison, it will intensify alienation. If it becomes a tool for better mediation, translation, education, and civic participation, it can reduce the very loneliness and confusion that make people susceptible to manipulation.

The central question is not whether AI is intelligent. It is whether it makes human life more inhabitable.

That is a more civic way to talk about technology. It asks whether systems help people feel seen without feeling incited, informed without being inflamed, and empowered without being weaponized.

A practical framework: four escalators and four brakes

To connect these worlds in a usable way, think in terms of escalators and brakes.

The four escalators

Distorted perception
People think violence is more common, more accepted, or more imminent than it really is.
Moral injury without repair
Grievance hardens into identity, and identity starts to require an enemy.
Algorithmic amplification
Digital systems reward the most polarizing content and intensify emotional contagion.
Institutional humiliation
Heavy-handed authority confirms the story that the system is hostile and illegitimate.

The four brakes

Procedural justice
Even when people lose, they are less likely to radicalize if they believe the process was fair.
Alternative belonging
People need groups that provide meaning, intensity, and purpose without requiring violence.
Early detection
Communities, schools, families, and platforms need tools to notice warning signals before action.
Governance of incentives
Both law and technology must stop rewarding outrage as if it were insight.

This framework matters because it breaks the false binary between individual pathology and structural causation. The truth is that violence is usually produced by the interaction of personal vulnerability and system design. Some people are more susceptible than others, but no society gets to pretend the surrounding environment is neutral.

Key Takeaways

Do not confuse low support for violence with low risk of violence.
Most people reject violence, but distorted perception can still make violence seem normal and imminent.
Treat grievance as a design problem, not only a belief problem.
People are often pulled by humiliation, status loss, belonging needs, and moral outrage, not just ideology.
Avoid responses that create “tyrannical peace.”
Heavy-handed control can reduce immediate disorder while increasing long-term resentment and escalation.
Build early warning systems that are social, not only technical.
Teachers, families, coworkers, and communities can notice leaked intent and behavioral changes before crisis.
Ask whether your systems reward outrage.
Whether you are designing policy, moderation, law enforcement practice, or AI tools, remove incentives that turn attention into aggression.

The deepest connection: both democracy and AI are trust systems

The final insight is simple but easy to miss. Political violence destroys trust in institutions. AI, if badly governed, can also destroy trust, not by attacking democracy directly, but by corroding the shared reality on which democracy depends. When people no longer trust that institutions are fair, that opponents are human, or that information is not being weaponized against them, the path to coercion gets shorter.

So the real common task is not merely to prevent violence or regulate AI. It is to protect the conditions under which disagreement remains nonviolent. That means more than punishing extremes after they emerge. It means building environments where grievance can be expressed without being amplified into apocalypse, where authority can be exercised without humiliation, and where technology serves human deliberation instead of replacing it.

The most important question is not, “How do we stop the next act of violence?” It is, “What kind of world makes violent answers feel plausible in the first place?” Once you ask that, AI regulation, democratic renewal, and violence prevention stop looking like separate agendas. They become different names for the same civilizational project: designing systems that keep people inside the human conversation.